Category: ComputerVision OpenCV

import cv2
import numpy as np

img = cv2.imread('out/result.jpg') 

gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# cv2.imwrite("out/gray.jpg", img_gray)

# 二値化
# threshold = 180
# _, img_th = cv2.threshold(gray, threshold, 255, cv2.THRESH_BINARY)
# cv2.imwrite("out/th.jpg", img_th)

# 反転
img_bw = cv2.bitwise_not(gray)
cv2.imwrite("out/bw.jpg", img_bw)

kernel = np.ones((3, 3), np.uint8)

img_ero = cv2.erode(img_bw, kernel, iterations=1)
cv2.imwrite("out/ero.jpg", img_ero)

モルフォジー前

モルフォジー後

iterationsの値を大きくするほど、白の値が狭くなる

反対にcv2.dilate(img_bw, kernel, iterations = 1) として白を白を膨張する事もできる

おおおおおおお
Sugeeeeeeeeeeee

画像を用意します

import cv2

image = cv2.imread("img/sweets.jpg")

gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)

edges = cv2.Canny(gray, 50, 200)

contours, hierarchy = cv2.findContours(edges, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_NONE)

count_objects_image = len(contours)

print(count_objects_image)

$ python3 main.py
232

illustratorなどで作った画像なら正確に抽出できるが、
リアルの写真だと上手く検出できないな…

$ sudo apt-get update
$ sudo apt-get install libopencv-dev python3-opencv
$ python3
Python 3.8.10 (default, Sep 28 2021, 16:10:42)
[GCC 9.3.0] on linux
Type “help”, “copyright”, “credits” or “license” for more information.
>>> import cv2
>>> cv2.__version__
‘4.2.0’

$ sudo apt-get install python3-pip
$ pip3 install pillow

import cv2

img = cv2.imread("src/test.jpg")

imgG = img.mean(axis=2)

from PIL import Image
imgP = Image.fromarray(imgG)
imgP = imgP.convert('RGB')
imgP.save('src/test2.jpg')

$ wget https://github.com/opencv/opencv/blob/master/data/haarcascades/haarcascade_frontalface_alt.xml

### face-detection
$ wget https://raw.githubusercontent.com/opencv/opencv/master/data/haarcascades/haarcascade_frontalface_default.xml

import cv2
 
cascade_file = "haarcascade_frontalface_default.xml"
cascade = cv2.CascadeClassifier(cascade_file)
 
img = cv2.imread("src/face.jpg")
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
 
face_list = cascade.detectMultiScale(img_gray, minSize=(50,50))
 
if len(face_list) == 0:
    print("false")
    quit()
 
for (x,y,w,h) in face_list:
    print(" 顔の座標=", x, y, w, h)
    red = (0, 0, 255)
    cv2.rectangle(img, (x, y), (x+w, y+h), red, thickness=5)
 
cv2.imwrite("src/face-detect.png", img)

マスク画像

マスクでも行けるのは行けるのか

2つ画像を用意します。
※プロントで作業しています。ワイヤレスマウスUBS

import cv2
import numpy as np

img_last = cv2.imread(“a.jpg”)
img_last = cv2.cvtColor(img_last, cv2.COLOR_BGR2GRAY) # 白黒化
img_last = cv2.GaussianBlur(img_last, (9, 9), 0) # ぼかし
img_last = cv2.threshold(img_last, 100, 255, cv2.THRESH_BINARY)[1] # 画像の二値化

img = cv2.imread(“b.jpg”)
img_current = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img_current = cv2.GaussianBlur(img_current, (9, 9), 0)
img_current = cv2.threshold(img_current, 100, 255, cv2.THRESH_BINARY)[1]

img_diff = cv2.absdiff(img_last, img_current)
cnts = cv2.findContours(img_diff, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)[0]

for pt in cnts:
x, y, w, h = cv2.boundingRect(pt)
if w < 30: continue cv2.rectangle(img, (x, y), (x+w, y+h), (0, 255, 0), 2) cv2.imwrite("result.png", img) [/python]

OK, Let’s Gooooooooooooooooooooo

import cv2
import numpy as np

img = cv2.imread("rose.jpg")

img[:, :, 0] = 0
img[:, :, 1] = 0

cv2.imwrite("result.png", img)

before

after

### HSVで赤っぽい部分の輪郭を表示

import cv2
import numpy as np

img = cv2.imread("rose.jpg")

hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV_FULL)
h = hsv[:, :, 0]
s = hsv[:, :, 1]
v = hsv[:, :, 2]

img2 = np.zeros(h.shape, dtype=np.uint8)
img2[(h > 200)  & (s > 100)] = 255

cv2.imwrite("result.png", img2)

ほう